Title: The impact of Four Factors on a basketball team success: An approach with model-based recursive partitioning
Authors: Manlio Migliorati - University of Brescia (Italy) [presenting]
Marica Manisera - University of Brescia (Italy)
Paola Zuccolotto - University of Brescia (Italy)
Abstract: According to some basketball experts, statistics are killing basketball. In our opinion, they are right, if statistics reduce the game to numbers that are not truly able to describe it. Instead, sound statistical methods start from those statistics as the input data, and appropriately elaborate and transform them into useful information to support technical experts. In the last decades, publications on statistics in basketball have multiplied and tried to answer different research questions: forecasting the outcomes of a game, analysing players performance, identifying optimal game strategies. We study the evolution of the weight of the Oliver Four Factors as determinants of the probability of winning a basketball game, using data from 19138 matches of 16 NBA regular seasons (from 2004-2005 to 2019-2020). Four Factors identify team strengths and weaknesses: shooting, turnovers, rebounding and free throws. Intending to investigate the role of each factor in determining a team success, we applied the MOB algorithm for model-based recursive partitioning that, instead of fitting one global model to the entire dataset, estimates local models on clusters of matches that are defined according to a learning algorithm based on recursive partitioning.